Create app.py

app.py

@@ -0,0 +1,56 @@
+import json
+import numpy as np
+from PIL import Image
+import tensorflow as tf
+from huggingface_hub import hf_hub_download
+import gradio as gr
+
+# — Model & labels download (runs once at startup) —
+MODEL_PATH = hf_hub_download(
+    repo_id="Adriana213/vgg16-fruit-classifier",
+    filename="vgg16-fruit-classifier.h5"
+)
+LABELS_PATH = hf_hub_download(
+    repo_id="Adriana213/vgg16-fruit-classifier",
+    filename="class_labels.json"
+)
+
+# — Load model & labels —
+model = tf.keras.models.load_model(MODEL_PATH)
+with open(LABELS_PATH, "r") as f:
+    id2label = json.load(f)
+
+# — Preprocessing helper —
+def preprocess(img: Image.Image) -> np.ndarray:
+    img = img.resize((100, 100))
+    arr = np.array(img)
+    if arr.ndim == 2:  # grayscale → RGB
+        arr = np.stack([arr]*3, axis=-1)
+    arr = arr.astype("float32")
+    # VGG16 preprocessing: subtract mean RGB
+    arr[..., 0] -= 123.68
+    arr[..., 1] -= 116.779
+    arr[..., 2] -= 103.939
+    return np.expand_dims(arr, 0)
+
+# — Prediction function —
+def classify_fruit(img: Image.Image):
+    img = img.convert("RGB")
+    x = preprocess(img)
+    preds = model.predict(x)
+    idx = int(np.argmax(preds, axis=1)[0])
+    label = id2label[str(idx)]
+    confidence = float(np.max(preds))
+    return {label: round(confidence, 4)}
+
+# — Gradio interface —
+demo = gr.Interface(
+    fn=classify_fruit,
+    inputs=gr.Image(type="pil", label="Upload Fruit Image"),
+    outputs=gr.Label(num_top_classes=5, label="Top Prediction"),
+    title="🍉 Fruit Classifier (131 types)",
+    description="Upload a photo of a fruit and get back its name with confidence."
+)
+
+if __name__ == "__main__":
+    demo.launch()